Available online at www.sciencedirect.

com

ScienceDirect
Resource-Efficient Technologies 2 (2016) S136–S152
www.elsevier.com/locate/reffit

Research paper
A comparative study of fractional order PIλ/PIλDμ tuning rules for stable
first order plus time delay processes
R. Ranganayakulu a, G. Uday Bhaskar Babu a,*, A. Seshagiri Rao a, Dipesh Shikchand Patle b
a
Department of Chemical Engineering, National Institute of Technology, Warangal Telangana 506004, India
b
School of Civil and Chemical Engineering (SCALE), VIT University, Vellore Tamilnadu 632014, India
Received 27 June 2016; received in revised form 11 November 2016; accepted 11 November 2016
Available online 20 December 2016

Abstract
Conventional PID tuning methods may not be sufficient to deal with complex processes of modern industry. For better control, fractional order
PIλDμ controller was introduced as the generalization of classical PID controller with the help of non-integer order (fractional order) calculus. The
fractional calculus uses integration and differentiation with a fractional order or complex order. The major advantage of fractional derivative is the
ability to inherit the nature of the processes. In general, the control loop includes both fractional order process model and fractional order controller.
However, the processes to be controlled are usually modeled as integer order models and controlled using fractional order controllers. But if the
plant model is obtained as fractional model, it is converted into integer order model by approximating the fractional terms using different
approximations proposed in the literature. With all the above mentioned advantages, several fractional order PIλ/PIλDμ tuning rules are proposed
in the literature for integer order systems and researchers are still proposing the new rules. The main aim of this paper is to compare fractional order
PI/PID tuning methods based on Integral of Absolute Error (IAE), Total Variation (TV) and Maximum Sensitivity (Ms). The main reason for
choosing fractional order PIλ/PIλDμ controllers is their additional degrees of freedom that result in better control performance. These tuning rules
were applied on several first order plus time delay processes subjected to step change in setpoint and disturbance.
Six recent tuning methods, three for fractional order PIλ and the remaining for fractional order PIλDμ, were considered. Finally, from the
simulation results the optimal tuning method is recommended based on the control objective of the process and the process dead time (L) to time
constant (T) ratio. It is observed that the performance of tuning methods vary with the nature of the process like lag dominant, balanced and delay
significant processes. The FOPTD processes were checked for robustness with increasing L/T ratio with respect to IAE, TV and Ms.
© 2016 Tomsk Polytechnic University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Fractional order; Tuning; Robustness; Sensitivity; Integral of absolute error; Maximum sensitivity

1. Introduction difficult to adjust PID parameters properly to meet the require-

ments because many industrial systems are often burdened with
The past six decades have seen the development of algo-
problems such as structural complexity, uncertainties, large
rithms for tuning PID controllers and their application in indus-
transportation lags and nonlinearities. The closed loop perfor-
try. The reason for the success of PID controllers is its lucid
mance of such systems can be enhanced using PIλDμ controller,
transfer function, the availability of good number of effective
providing additional degrees of freedom with fractional order
tuning rules and their acceptance by the industrial world based
of the integral λ and derivative μ. This flexibility allows design-
on minimum knowledge of the process to be controlled. The
ing of more robust systems, thus providing the precise control
most widely used controller in the industry is PID due to its
of real world processes. The strength of FOPID controller can
simple structure and auto tuning capability. Although there are
be attributed to its robustness, i.e., less sensitive to uncertainties
many rule based methods and analytical tuning methods, it is
in the process parameters and the ability to control both integer
order and fractional order dynamical systems. In the field of
automatic control, these advantages would lead to more precise
* Corresponding author. Department of Chemical Engineering, National
Institute of Technology, Warangal 506004, India. Tel.: +91 8332969404; fax: 91
and robust control of dynamical systems.
870 2459547. Better control is always a primary objective in the field of
E-mail address: udaybhaskar@nitw.ac.in (G. Uday Bhaskar Babu). control engineering and opens new boundaries for developing
http://dx.doi.org/10.1016/j.reffit.2016.11.009
2405-6537/© 2016 Tomsk Polytechnic University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer review under responsibility of Tomsk Polytechnic University.
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152
much better control algorithms. The fractional order propor-
tional integral derivative (FOPID or PIλDμ) controller has been
a topic of interest of researchers from academia and industry in
the last ten years. They are more flexible and provide improved
performance than conventional PID controllers due to the pres-
ence of five parameters to be tuned (rather than three in the case
of traditional PID). However, the tuning of FOPID controller
becomes complex. To address these problems, different design
and tuning algorithms have been proposed in the literature for
PIλ/ PIλDμ controllers.
A closed loop control system encounters different combina-
tions of plant and controllers while handling real world prob-
lems. They include the integer or fractional order of either plant
or controller or both. In practice, the plant models have been
obtained as integer order models and it is natural to consider the
fractional nature of the controller. The efforts of control engi-
neers and scientists lead to the development of fractional order
controller (PIλ/PIλDμ) tuning rules. Podlubny [1] had proposed
the fractional order PI and PID controllers that demonstrated
better response with integrator and differentiator rose to the
fractional powers λ and μ respectively. There are several other
PIλ and PIλDμ controllers in the literature [2–6] whose tuning
rules are developed using evolutionary algorithms by minimiz-
ing objective functions. Some heuristics based controllers are
also reported.
Many industrial processes are better approximated as first
order plus time delay (FOPTD) models. In fact, there are rules
to approximate higher order processes and processes with non-
linearity to fit to FOPTD models [7]. This paper is restricted to
the use of integer order FOPTD models due to the above men-
tioned reasons for controller design. This study determines the
optimal PIλ/PIλDμ tuning rules to calculate controller settings
for stable FOPTD processes. The work also suggests the suit-
able tuning method depending on the control objective. The
closed loop performance of two lag dominant FOPTD pro-
cesses is studied. Further, the variation of performance indices
and robustness of the controller are considered.
In the current work, the following six PIλ/PIλDμ tuning rules
were considered based on the review of recent tuning rules in
the literature. Initially, the FOPI tuning rules were studied
[8–10]. The FOPID tuning rules [11–13] were then chosen for
study as they reduce the effect of overshoot caused by integral
action in FOPI. The present work compares the tuning rules
with regard to performance metrics integral of absolute error
(IAE), total variation (TV) and maximum sensitivity (Ms) for
integer order FOPTD models.
2. Background
2.1. Block diagram of closed loop control system
The general feedback loop used for the control of single
input single output processes is shown in Fig. 1. The various
parameters in the loop are: ysp – set value, y – controlled
variable, u – manipulated variable and e(=ysp-y) – the error.
Another input variable which enters the process and affects the
controlled variable is the disturbance ‘d’. Gp(s) represents a
stable process which includes the dynamics of measuring
S137
Fig. 1. Feedback control loop with PIλ/PIλDμ controller.
element and final control element and is modeled by an FOPTD
transfer function:
K − Ls
G p (s ) = e (1)
Ts + 1
The PIλ/PIλDμ block contains the transfer function of con-
troller and is represented in ideal form or parallel form as
follows:
⎛ 1 ⎞ K
G c (s ) = K p ⎜1 + λ ⎟ = K p + λi or
⎝ Ts i ⎠ s
(2)
⎛ 1 ⎞ K
G c (s ) = K p ⎜1 + λ + Td sμ ⎟ = K p + λi + K d sμ
⎝ Ts ⎠ s
i
The series form of PIλDμ controller considered in the litera-
ture is:
⎛ Ts λ + 1⎞ ⎛ T sμ + 1 ⎞
Kp ⎜ i λ ⎟ ⎜ d ; N = 10Tμ −1 (3)
⎝ Ts i ⎠ ⎝ (Td N ) s + 1⎟⎠
Chen et al., Gude et al., and Bhambhani et al. used the
parallel form of FOPI controller [8–10]. Also, Valerio et al., and
Bayat used the parallel form of FOPID controller, whereas
Padula et al. used the series form of FOPID controller [11–13].
Series form and parallel forms of controller structures are used
according to the application. It is observed that the parallel
form of PID controller is more often used in academia and in
research while the series form of PID controller is used in
industrial controllers. The ultimate purpose of both forms of
controllers is the same and it is possible to convert controller
structure from one form to the other using associated relations.
2.2. Performance assessment of the controller
The generally used measures for controller performance are
integral error criteria: integral of squared error (ISE), integral
of absolute error (IAE) and integral of time weighted absolute
error (ITAE). In the current study, IAE criterion was chosen as
performance metric for the controller because the minimization
of IAE results in small overshoot and low settling time in the
closed loop response of the system when there is a step change
in set point or load disturbance. It is defined as:
∞
IAE = ∫ e ( t ) dt (4)
0
Another controller performance metric is the total variation
(TV) of the manipulated variable. Total variation is an indica-
tion of the smoothness of the manipulated variable. Total varia-
tion is defined as:
S138 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152

∞ 3.2. Gude et al. [9] method

TV = ∑ u i+1 − u i (5)
i=0 The tuning rules are developed by minimizing a perfor-
mance criteria Jv in frequency domain with a constraint on the
For effective and stable function of the closed loop control
maximum sensitivity. Jv measures the ability of the closed loop
system, it should not be sensitive to process parameter changes
system to handle disturbance inputs of low frequencies.
or model uncertainties. This is a major requirement for robust-
ness of the controller. The sensitivity function S is defined as: 1 1 G ( jω )
Jv = G (s ) S (s ) = max (8)
1 s ∞
ω jω 1 + L ( jω )
S ( jω ) = (6)
1 + G c ( jω ) G p ( jω )
3.3. Bhambhani et al. [10] method
This sensitivity function tells us how a closed loop system is
influenced by parameter variations in the process. The These tuning rules were proposed based on the minimization
maximum value of the sensitivity function is a performance of two controller performance indices: Jitter margin and ITAE.
metric for robustness analysis: Multi-objective optimization algorithm is used to calculate the
optimum values of the tuning parameters.
Ms = max S ( jω ) (7)
0<ω <∞ 1 1 + G ( jω ) C ( jω )
Jitter margin = ; δ max = min (9)
The value of maximum sensitivity Ms can also be found δ max 0 <ω <∞ jωG ( jω ) C ( jω )
from the Nyquist stability test of closed-loop system. For the
∞
stability of closed loop system, the Nyquist curve should not
ITAE = ∫ t e ( t ) dt (10)
enclose the point (−1, j0). The maximum sensitivity is the
0
reciprocal of shortest distance between Nyquist curve and the
point (−1, j0). If the value of Ms increases, the system becomes
3.4. Valerio et al. [11] method
less robust to modeling uncertainties. The typical range of
maximum sensitivity should be 1 to 2 to ensure robust perfor- These rules are applicable for processes whose unit step
mance of the controller. response is of s-shaped. The advantage of this method is that the
controller design does not require the plant data. Two sets of
3. PIλ/PIλDμ tuning rules tuning rules were proposed: first set of tuning rules are appli-
cable for processes with 0.1 ≤ T ≤ 50and L ≤ 2, and second set
This section describes six tuning methods considered in this for 0.1 ≤ T ≤ 50 and L ≤ 0.5. A numerical optimization algo-
work and the formulas for controller settings of all the tuning rithm namely Nelder–Mead’s simplex method is implemented
methods are listed in Tables 1 and 2. in MATLAB to minimize the condition:

3.1. Chen et al. [8] method G ( jω cg ) C ( jω cg ) = 0dB (11)

This method proposed tuning parameters of the controller by These rules are easier to apply but perform worse than ana-
optimizing the regulatory response with a constraint on the lytical tuning methods. Fine tuning is often required for con-
maximum load disturbance-to-output sensitivity Ms. trollers designed based on these tuning rules.

Table 1
Fractional order PI controller tuning rules.
Tuning rule Controller settings

⎧1.1 if τ ≥ 0.6
1 ⎛ 0.2978 ⎞ ⎛ ⎞ ⎪1.0 0.4 ≤ τ < 0.6
Kp = ⎜ Ti = T ⎜ 2
0.8578 ⎪ if
Chen et al. [8] ⎟ λ=⎨
K ⎝ τ + 0.000307 ⎠ ⎝ τ − 3.402τ + 2.405 ⎟⎠ 0.1 ≤ τ < 0.4
⎪0.9 if
⎩⎪0.7 if τ < 0.1
Gude et al. [9] f ( τ ) = aτ b + c a b c τ
KK p 0.2154 −1.169 −0.1592 0<τ<1
aK −0.4645 0.3182 0.5795 0 < τ < 0.25
3.271 5.75 0.28 0.25 < τ < 1
Ti L 9.242 −0.1966 −9.171 0<τ1
Ti T 5.479 0.8154 −0.03853 0 < τ < 0.3
6.06 7.066 1.18 0.3 < τ < 1
λ 1.12 0<τ<1
0.2T 0.25 0.19833
Bhambhani et al. [10] Kp = + 0.16 Ki = + + 0.09 λ = τ−0.04 L + 1.2399
L TL L
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152 S139

Table 2
Fractional order PID controller tuning rules.
Valerio et al. [11]
First set of tuning rules
For 0.1 ≤ T ≤ 5 and L ≤ 2 For 5 ≤ T ≤ 50 and L ≤ 2
P I λ D μ P I λ D μ

1 −0.0048 0.3254 1.5766 0.0662 0.8736 2.1187 −0.5201 1.0645 1.1421 1.29
L 0.2664 0.2478 −0.2098 −0.2528 0.2746 −3.5207 2.6643 −0.3268 −1.3707 −0.5371
T 0.4982 0.1429 −0.1313 0.1081 0.1489 −0.1563 0.3453 −0.0229 0.0357 −0.0381
L2 0.0232 −0.133 0.0713 0.0702 −0.1557 1.5827 −1.0944 0.2018 05552 0.2208
T2 −0.072 0.0258 0.0016 0.0328 −0.025 0.0025 0.0002 0.0003 −0.0002 0.0007
LT −0.0348 −0.0171 0.0114 0.2202 −0.0323 0.1824 −0.1054 0.0028 0.263 −0.0014
Second set of tuning rules
For 0.1 ≤ T ≤ 50and L ≤ 0.5
P I λ D μ
1 −1.0574 0.6014 1.1851 0.8793 0.2778
L 24.542 0.4025 −0.3464 −15.0846 −2.1522
T 0.3544 0.7921 −0.0492 −0.0771 0.0675
L2 −46.7325 −0.4508 1.7317 28.0388 2.4387
T2 −0.0021 0.0018 0.0006 −0.0000 −0.0013
LT −0.3106 −1.205 0.038 1.6711 0.0021
Bayat [12]
Servo control Regulatory control
1 ⎡ 0.3663τ + 0.8856 ⎤ 1 ⎡ 0.2709τ + 0.566 ⎤
Kp
K ⎢⎣ τ + 0.000792 ⎥⎦ K ⎢⎣ τ − 0.0364 ⎥⎦
Ti T (0.3827τ + 0.9354) T (1.252τ 0.5555 − 0.05696 )
Td T (0.5036τ 0.7152 − 0.07974 ) T (0.3425τ + 0.02753)
λ 1 1
μ 1.095−0.03625τ 2.503 − 1.368τ 0.04705
Padula et al. [13]
For Ms = 1.4
1
For servo control For regulatory control Kp =
K
( aτ b + c )
f (τ) a b c a b c
⎛ L b ⎞
Kp 0.6503 −0.9166 −0.6741 0.2776 −1.095 −0.1426 Ti = Tλ ⎜ a ⎛⎜ ⎞⎟ + c⎟
Ti 0.04701 −0.2611 0.9276 0.6241 0.5573 0.0442 ⎝ ⎝ T⎠ ⎠
Td 0.3563 1.2 0.0003108 0.4793 0.7469 −0.02393 ⎛ L b ⎞
1 1 Td = Tμ ⎜ a ⎛⎜ ⎞⎟ + c⎟
λ
⎝ ⎝ T⎠ ⎠
μ 1.1 for τ < 0.1 1.0 for τ < 0.1
1.2 for τ ≥ 0.1 1.1 for 0.1 ≤ τ < 0.4
1.2 for τ ≥ 0.4

3.5. Bayat [12] method optimal tuning rules were obtained by interpolating the optimal
values found for controller parameters of different processes
These tuning rules were proposed based on the dimensional
with different dead time τ. The structure of the tuning param-
analysis and are applicable for processes modeled as FOPTD.
eters devised is given as follows:
The only limitation is that the normalized dead time should be
between 0.1 and 3.5. Tuning rules for command following and 1
disturbance rejection were proposed exclusively to minimize
Kp =
K
( aτ b + c ) (12)
ISE performance index.
⎛ ⎛ L⎞ b ⎞
Ti = Tλ ⎜ a ⎜ ⎟ + c⎟ (13)
3.6. Padula et al. method ⎝ ⎝ T⎠ ⎠
Padula et al. [13] proposed FOPID tuning rules explicitly for
servo control and regulatory control of closed loop system by ⎛ ⎛ L⎞ b ⎞
Td = Tμ ⎜ a ⎜ ⎟ + c⎟ (14)
minimizing integrated absolute error (IAE) with a constraint on ⎝ ⎝ T⎠ ⎠
the maximum sensitivity Ms. The series form of FOPID con-
troller was considered for proposing the tuning rules. The
4. Simulation and comparison study
tuning rules were developed by considering several processes
with various dead time values. The values of the tuning param- A stable FOPTD model is considered for the current study
eters had been found by using genetic algorithm. Finally, the [14]:
S140 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152
1 − Ls
G p (s ) = e (15)
s +1
For analogy, the performance of controller tuning methods is
compared for variations in load and set point when they
undergo a step change of unit magnitude. L/T ratio is a signifi-
cant factor which affects the controller performance and sensi-
tivity of the feedback control system. The effect of L/T ratio on
different tuning methods was studied by varying time delay L
so that the ratio L/T varies from 0.1 to 2 covering lag dominant,
balanced and delay significant processes. The simulations were
carried out on different FOPTD processes. The main reason for
varying the L/T ratio is that it affects the robustness of control-
ler and performance of the closed loop system. For each varia-
tion of L/T, new controller settings are calculated and closed
loop response (both servo and regulatory) is observed, thus
recording IAE, TV and Ms. The trends of the performance
indices IAE, TV and Ms are observed for the entire range of L/T
ratio. Finally, the behavior of IAE, TV and Ms is observed after
plotting them against L/T ratio. Also, the performance of the
fractional controllers was observed by adjusting the fractional
tuning parameters λ and μ. All the simulations are carried out
using MATLAB and Simulink.
4.1. Case study: liquid level system
To verify the recommended tuning rules, the case study of
liquid level system modeled as FOPTD [15] is considered. The
transfer function of liquid level system modeled as FOPTD
system is:
3.13
G p (s ) = e −50s (16)
433.33s + 1
Fig. 2. Closed loop responses of lag dominant process for set-point change and L/T = 0.3 (a) FOPI, (b) FOPID.
The value of L/T = 0.1034 < 1 indicates that the system is a
lag dominant process and the behavior of performance indices
are observed for the variation of L/T ratio.
5. Results and discussion
The simulation studies on servo control and regulatory
control problems are presented for two stable FOPTD pro-
cesses. The IAE, TV and Ms values are expected to be minimum
for better performance of the controller. The servo response of
the closed loop system and the corresponding FOPI and FOPID
controller responses are shown in Fig. 2 for lag dominant
process. It is evident from the response that the Gude et al. [9]
method provides better response. The Chen et al. [8] method
also provides a good response but left with an offset in the
response. This is due to the low value of fractional order of the
integral action, i.e., 0.9 as the normalized dead time is 0.2307.
The Bhambhani et al. [10] method is giving an oscillatory
response. It is clear from the response of the closed loop system
with FOPID controller that the Padula et al. [13] method results
in a stable and offset free response without any overshoot while
the other two methods produce an oscillatory response. There
appears an offset in the response when the controller is
designed with Valerio et al. [11] tuning rules because the frac-
tional order λ of the integral is 0.8281. Further, the performance
of the controllers was observed by varying the fractional param-
eters λ and μ over a possible range values, starting with a value
that gives a stable response (Fig. 3).
5.1. Servo control
5.1.1. Servo control of lag dominant process
The stable FOPTD model considered for the study becomes
lag dominant for any value of time delay, between 0.1 and 0.9.
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152
Fig. 3. (a) Legend for Figs 4–6,10–12,16–17 and 20–21. (b) Legend for
Figs 7–9,13–15,18–19 and 22–23.
Here, it is chosen as 0.3 so that the ratio L/T becomes 0.3. Fig. 4
shows the variation of performance metrics IAE and TV of the
FOPI controller as L/T varied from 0.1 to 2. Fig. 5 shows
the robustness Ms of the controller for servo control problem.
The Chen et al. [8] method gives lower values of IAE and TV
whereas the Gude et al. [9] method is robust to model uncer-
tainties compared to the other two methods. Both the methods
are robust to model parameter uncertainties because the
maximum sensitivity lies in the range of 1.2 to 2. The controller
effort in terms of total variation is the same in case of both
methods. The Bhambhani et al. [10] method fails to be robust to
model uncertainties as it gives much higher values of Ms com-
pared to the other two methods and also the higher values of
IAE. Fig. 5 also clarifies that the Gude et al. [9] method per-
forms better compared to the other two methods even if the
tuning parameter λ is varied over a range of values keeping
others unaltered. The Chen et al. [8] method can be a second
choice for good system performance if offset is tolerated
(Fig. 6).
Fig. 4. Increasing L/T ratio versus performance metrics of FOPI controller for step set point changes of unit magnitude for the three tuning methods.
S141
The performance of FOPID controller is presented in
Figs 7–9 for variation of controllability index L/T. Lower values
of IAE and TV are observed with the Padula et al. [13] method.
The Valerio et al. [11] method shows significantly higher values
compared to Padula et al. [13] method. The Bayat [12] method
shows poor performance both in terms of IAE and TV. The
robustness of the controller presented in Fig. 8 indicates that the
Padula et al. [13] method is robust to model uncertainties.
FOPID controller performance presented in Fig. 9 with the
variation of tuning parameter λ confirms that the Padula et al.
[13] method is superior in terms of IAE, TV and Ms. The
Valerio et al. [11] method can be a second choice but its appli-
cation is limited to the systems with s-shaped step response;
time delay must be less than or equal to 2 and time constant
should be less than or equal to 50.
5.2. Regulatory control
5.2.1. Regulatory control of lag dominant process
Figures 10 and 11 show the variation of performance metrics
IAE, TV of the FOPI controller and sensitivity for regulatory
control problem. It is observed that the Gude et al. [9] method
results in lower values of TV whereas the Chen et al. [8] method
gives lower values of IAE but with an offset in the response as
long as the fractional order of the integrator is below 1. The
graphs illustrated in Fig. 11 reveal that the Gude et al. [9]
method is insensitive to the model parameter variations and the
Chen et al. [8] method too as its Ms values are also in the range
of 1.2 to 2. Finally, the Gude et al. [9] method is recommended
S142 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152

Fig. 5. Robustness of FOPI controller for step set point changes of unit magnitude with increasing L/T ratio for the three tuning methods.

Fig. 6. FOPI controller performance for step set point changes of unit magnitude with L/T ratio of 0.3 for the three tuning methods.
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152 S143

Fig. 7. Increasing L/T ratio versus performance metrics of FOPID controller for step set point changes of unit magnitude for the three tuning methods.

Fig. 8. Robustness of FOPID controller for step set point changes of unit magnitude with increasing L/T ratio for the three tuning methods.
S144 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152

Fig. 9. FOPID controller performance for step set point changes of unit magnitude with L/T ratio of 0.3 for the three tuning methods.

Fig. 10. Increasing L/T ratio versus performance metrics of FOPI controller for step disturbance changes of unit magnitude for the three tuning methods.
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152 S145

Fig. 11. Robustness of FOPI controller for step disturbance changes of unit magnitude with increasing L/T ratio for the three tuning methods.

for better closed loop performance except for a bit higher IAE ity index L/T, the plots of performance metrics IAE, TV and Ms
values. Fig. 12 presents the performance of FOPI controller with respect to increasing L/T ratio for FOPI controller tuning
with the variation in tuning parameter λ. It ensures the overall methods and FOPID controller tuning methods were illustrated
performance of the closed loop system with the Gude et al. [9] for unit step type change in set point and disturbance. It is
method of controller settings.
From Fig. 13, it is observed that the Padula et al. [13] method
gives lower values of IAE and TV followed by the Valerio et al.
Table 3
[11] method. Controller effort is almost the same in the case of Recommendation of tuning rules for FOPI controller.
all three methods. In terms of Ms value, the Padula et al. [13] Input Lag dominant
method is giving a value of 1.4 throughout the scale of L/T ratio
Servo control Unit step 1. Gude et al. [9]
as it was designed to be. The Valerio et al. [11] method is giving 2. Chen et al. [8]
Ms values beyond 2 when the process becomes delay signifi- Regulatory control Unit step 1. Gude et al. [9]
cant. The Bayat [12] method is far from the requirements in 2. Chen et al. [8]
terms of IAE and Ms. The result shown in Fig. 14 tells that
FOPID controller is robust with the controller settings calcu-
lated using the Padula et al. [13] method. The performance of Table 4
FOPID controller with the variation of λ illustrated in Fig. 15 Recommendation of tuning rules for FOPID controller.
clarifies that the Padula et al. [13] method is a better choice for Input Lag dominant
robust performance of the controller and stable response of the Servo control Unit step 1. Padula et al. [13]
FOPTD system. Regulatory control Unit step 1. Padula et al. [13]
2. Valerio et al. [11]
5.3. Case study: liquid level system
It is easy to identify the optimal tuning rules of FOPI and Table 5
FOPID controllers with the known value of L/T and with the Optimal FOPI tuning rules identified for level control system.
recommendations given in Tables 3 and 4 for set point tracking Input Lag dominant
and disturbance rejection problems. The optimal tuning rules Servo control Unit step 1. Chen et al. [8]
identified are summarized in Tables 5 and 6. 2. Gude et al. [9]
Figures 16–23 illustrate the simulation results obtained for Regulatory control Unit step 1. Chen et al. [8]
2. Gude et al. [9]
the level control system. To identify the effects of controllabil-
S146 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152

Fig. 12. FOPI controller performance for step disturbance changes of unit magnitude with L/T ratio of 0.3 for the three tuning methods.

Fig. 13. Increasing L/T ratio versus performance metrics of FOPID controller for step disturbance changes of unit magnitude for the three tuning methods.
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152 S147

Fig. 14. Robustness of FOPID controller for step disturbance changes of unit magnitude with increasing L/T ratio for the three tuning methods.

Fig. 15. FOPID controller performance for step disturbance changes of unit magnitude with L/T ratio of 0.3 for the three tuning methods.
S148 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152
Table 6
Optimal FOPID tuning rules identified for level control system.
Inputs Lag dominant
Servo control Unit step 1. Padula et al. [13]
Regulatory control Unit step 1. Padula et al. [13]
evident from Figs 16–17 and Figs 20–21 that the Chen et al. [8]
method of tuning FOPI controller shows minimum values of
IAE and Ms for servo control and regulatory control problems.
The only limitation observed with the Chen et al. [8] method is
that it produces an offset in the response if the fractional order
λ is less than 1. The simulation results do not include the
Bhambhani et al. [10] method of FOPI controller response. This
is because the controller settings result in the negative value of
tuning parameter λ which causes improper transfer function of
the process model. It is apparent from Figs 18–19 and
Figs 22–23 that the Padula et al. [13] method is the superior
choice for tuning FOPID controller for servo control and regu-
latory control problems. The optimal tuning rules recom-
mended in Tables 3 and 4 match with the results obtained which
are shown in Tables 5 and 6. The Valerio et al. [11] method
cannot be used for this case study as the process model param-
eters L and T are higher than the designed values.
Better closed loop response can be achieved when FOPI
controller is tuned by the Chen et al. [8] and Gude et al. [9]
tuning rules. But the application of Chen et al. [8] tuning rules
for some lag dominant processes may result in an overshoot in
the response. The Gude et al. [9] tuning rules are suggested for
Fig. 16. Increasing L/T ratio versus performance metrics of FOPI controller for step set point changes of unit magnitude for the case study.
such processes. The high rise time, overshoot and settling time
make the Bhambhani et al. [10] tuning rules least preferred for
tuning the FOPI controller. The application of FOPID tuning
rules by Valerio et al. [11] is limited to the processes with
s-shaped step response. A further tuning is often required for
the controller tuned by the Valerio et al. [11] tuning rules. The
all time choice for tuning the FOPID controller is the Padula
et al. [13] tuning method because the closed loop response had
witnessed lower values of performance indices. The Bayat [12]
tuning rules result in higher values of performance indices and
hence not a choice for tuning the FOPID controller but reject
disturbance effectively from some of the processes.
6. Conclusions
This paper recommends optimal PIλ/PIλDμ tuning rules for
servo and regulatory control of processes in the industries when
there is step change in the input variables (set point, distur-
bance) of process loop. For servo control problem, the Gude
et al. [9] method is recommended followed by the Chen et al.
[8] method. An offset occurs if the Chen et al. [8] method is
used for servo control of lag dominant process if λ is less than
1. Hence, the Gude et al. [9] method is a primary choice for lag
dominant processes. For more robust setting, one needs a com-
promise between the Chen et al. [8] method and the Gude et al.
[9] method. The Padula et al. [13] PIλDμ tuning method is
recommended for servo control problem of all types of indus-
trial processes. For regulatory control, the Chen et al. [8]
method and the Gude et al. [9] method can alternatively be used
with a line between IAE and Ms. In case of PIλDμ tuning
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152 S149

Fig. 17. Robustness of FOPI controller for step set point changes of unit magnitude with increasing L/T ratio for the case study.

Fig. 18. Increasing L/T ratio versus performance metrics of FOPID controller for step set point changes of unit magnitude for the case study.
S150 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152

Fig. 19. Robustness of FOPID controller for step set point changes of unit magnitude with increasing L/T ratio for the case study.

Fig. 20. Increasing L/T ratio versus performance metrics of FOPI controller for step disturbance changes of unit magnitude for the case study.
 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152 S151

Fig. 21. Robustness of FOPI controller for step disturbance changes of unit magnitude with increasing L/T ratio for the case study.

Fig. 22. Increasing L/T ratio versus performance metrics of FOPID controller for step disturbance changes of unit magnitude for the case study.
S152 R. Ranganayakulu et al. / Resource-Efficient Technologies 2 (2016) S136–S152
Fig. 23. Robustness of FOPID controller for step disturbance changes of unit magnitude with increasing L/T ratio for the case study.
methods, the Padula et al. [13] method is recommended for fast
response and robustness of all kinds of industrial processes.
References
[1] I. Podlubny, Fractional-order systems and PIλDμ controllers, IEEE Trans.
Automat. Contr. 44 (1) (1999) 208–214.
[2] P. Shah, S. Agashe, Review of fractional PID controller, Mechatronics 38
(2016) 29–41.
[3] Y. Luo, Y.Q. Chen, Fractional Order Motion Control, first ed., John Wiley
and Sons, Chichester, UK, 2013.
[4] R.D. Keyser, C.I. Muresan, C.M. Ionescu, A novel auto-tuning method for
fractional order PI/PD controllers, ISA Trans. 62 (2016) 268–275.
[5] A.A. Kesarkar, N. Selvaganesan, Tuning of optimal fractional order PID
controller using an artificial bee colony algorithm, Syst. Sci. Control Eng.
3 (2015) 99–105.
[6] F. Padula, A. Visioli, On the fragility of fractional-order PID controllers
for FOPDT processes, ISA Trans. 60 (2016) 228–243.
[7] D.E. Seborg, T.F. Edgar, D.A. Mellichamp, Process Dynamics and
Control, second ed., John Wiley and Sons, New York, 2004.
[8] Y. Chen, T. Bhaskaran, D. Xue, Practical tuning rule development for
fractional order proportional and integral controllers, J. Comput.
Nonlinear Dyn. 3 (2) (2008) 214031–214038.
[9] J.J. Gude, E. Kahoraho, Simple tuning rules for fractional PI controllers.
IEEE Conference on Emerging Technologies & Factory Automation,
(2009) 4910–4915.
[10] V. Bhambhani, Y.Q. Chen, D. Xue, Optimal Fractional Order Proportional
Integral Controller for Varying Time-Delay Systems. 17th IFAC World
Congress, Seoul, Korea, July 6–11 (2008).
[11] D. Valerio, J.S.D. Costa, Tuning of fractional PID controllers with
Ziegler–Nichols-type rules, Signal Processing 86 (2006) 2771–2784.
[12] F.M. Bayat, General rules for optimal tuning the PIλDμ controllers with
application to first-order plus time delay processes, Can. J. Chem. Eng.
9999 (2012) 1–11.
[13] F. Padula, A. Visioli, Tuning rules for optimal PID and fractional order
PID controllers, J. Process Control 21 (2011) 69–81.
[14] D. Chen, D.E. Seborg, PI/PID controller design based on direct synthesis
and disturbance rejection, Ind. Eng. Chem. Res. 41 (2002) 4807–4822.
[15] C.A. Monje, B.M. Vinagre, V. Feliu, Y.Q. Chen, Tuning and auto-tuning of
fractional order controllers for industry applications, Control Eng. Pract.
16 (2008) 798–812.